Embedded CMAC learning controller for scan-tracking measurement in copying manufacture: results of a case study

Surface scan-tracking measurement is one of the key technologies in copying manufacture. In conventional scan-tracking measuring processes of irregular surface, the model surface geometric shape and the friction between the probe and model surface are two main factors that can seriously affect the measuring accuracy and efficiency. In order to reduce the impact of these factors and improve measurement efficiency while maintaining measurement accuracy, this paper presents a novel embedded cerebellar modular articulation controller (CMAC) learning controller for scan-tracking measurement in copying manufacture. New approaches to model surface features (including geometric feature and friction feature) identification and quantification are given specifically. Conventional scan-tracking control law is improved by taking into account the impact of model surface feature, and it is combined with CMAC neural network so that it can automatically predict the surface features and adjust the scan-tracking velocity in advance. Thus, high measuring efficiency can be obtained by accelerating scan speed in smooth areas of model surface and decelerating prior to scanning the surface feature cusp regions. Working with a commercial open CNC system, the design steps, integration process, and results of applying the embedded CMAC learning controller were described in detail through the examples of real measurement. Actual industrial tests show a higher measurement efficiency which demonstrates the effectiveness of proposed control strategy for scan-tracking measurement.